Using NGS to detect CNVs in familial hypercholesterolemia
- 1. Use of next-generation sequencing to detect copy number variants in the molecular diagnosis of familial hypercholesterolemia Michael Iacocca Research Trainee Blackburn Cardiovascular Genetics Laboratory Robarts Research Institute, Western University, Canada Supervisor: Dr. Robert Hegele, MD
- 2. Overview • What is familial hypercholesterolemia (FH) and why is it important? • What are the causes of FH and how is it being currently diagnosed at the molecular level? • Method: How can the molecular diagnosis be potentially improved? • What are the implications of this method? • How can this method be further applied?
- 3. Familial Hypercholesterolemia (FH) • Genetically determined extreme LDL cholesterol (LDL-C plasma concentration >95th percentile for age/sex) • Autosomal dominant inheritance • Heterozygous FH: Prevalence of ~1 in 250 (Akioyamen LE et al. BMJ Open. 2017) -most common monogenic disorder worldwide • < 10 % diagnosed globally • Early onset atherosclerosis causing CVD - ↑ risk of MI, stroke • Effectively lowered LDL-C, ~ normal life expectancy
- 4. Familial Hypercholesterolemia (FH) • DNA testing a central part of diagnosis worldwide (ClinVar at NCBI, accessed Dec 2017) Current method: 1) Targeted NGS panel - small-scale variants - LDLR, APOB, PCSK9 2) MLPA - large-scale CNVs (deletion/duplications of one or more whole exons) - LDLR • LDLR: loss-of-function variants • APOB: specific protein- altering variants • PCSK9: gain-of-function variants
- 5. Iacocca MA and Hegele RA. Curr Opin Lipidol. 2018 Unique LDLR CNVs identified in FH patients worldwide
- 6. Objective • To determine the potential of applying bioinformatics to existing NGS data to accurately detect CNVs in LDLR, thus removing the need for secondary MLPA analysis
- 7. Methods Study subjects • 388 individuals from Canada with a clinical diagnosis of at least ‘probable’ FH per the DLCN criteria Next-generation sequencing (NGS) • LipidSeq • 73 genes, including LDLR, APOB, PCSK9 and LDLRAP1, APOE, STAP1, ABCG5, ABCG8, LIPA CNV analysis by MLPA • Multiplex PCR method • Assay of promoter and all 18 exons in LDLR CNV analysis by NGS data • Bioinformatics applied to existing NGS data • VarSeq CNV Caller: Depth of coverage analysis
- 8. Methods: NGS Panel LipidSeq Panel NGS • 73 lipid metabolism-related genes, including all FH-associated genes LDLR, APOB, PCSK9 and LDLRAP1, APOE, STAP1, ABCG5, ABCG8, LIPA - All exons, 150 bp at intron/exon boundaries, ~250 bp of 5’UTR - 178 SNP loci • Library prep: Nextera Rapid Capture Custom Enrichment kit (Illumina) • Platform: MiSeq (Illumina) – 2 x 150 bp paired-end chemistry • Avg. 300-fold coverage per base Johansen CT et al. J Lipid Research. 2014 Hegele RA et al. Curr Opin Lipidol. 2015
- 9. Dilliott AA et al. J Vis Exp. 2018 Methods: CLC Genomics Workbench .FASTQ file .BED file NGS MiSeq Output
- 10. VarSeq CNV Caller Requirements • .BAM file • .VCF file 1) Patient sample 2) Matched reference controls (N= 30 to 50) • .BAM file • .VCF file 3) .BED file
- 11. Results MLPA NGS Type Region Detection Avg. Ratio Avg. Z-score Het. Deletion Promoter-Exon 1 (n=22) Yes 0.51 -6.2 Het. Deletion Promoter-Exon 2 (n=2) Yes 0.57 -5.6 Het. Deletion Promoter-Exon 6 Yes 0.54 -7.4 Het. Deletion Exons 2-3 Yes 0.56 -6.7 Het. Deletion Exons 2-6 Yes 0.54 -9.7 Duplication Exons 2-6 Yes 1.38 11.8 Het. Deletion Exons 3-6 Yes 0.53 -9.7 Het. Deletion Exons 5-6 Yes 0.54 -14.7 Duplication Exon 7 Yes 1.47 7.3 Duplication Exons 11-12 Yes 1.86 12.7 Het. Deletion Exons 11-12 Yes 0.54 -7.8 Het. Deletion Exons 13-14 Yes 0.51 -15.9 Het. Deletion Exons 13-15 Yes 0.65 -8.7 Het. Deletion Exons 16-18 Yes 0.53 -9.9 Het. Deletion Exons 17-18 (n=2) Yes 0.53 -9.3 CNVs in LDLR detected by MLPA • 38 of 388 (9.8%) FH patients were CNV positive Iacocca MA, Wang J, et al. J Lipid Res. 2017
- 12. Ex) VarSeq NGS data output: LDLR Exons 2- 6 heterozygous deletion
- 13. Ex) VarSeq NGS data output: LDLR Exons 11-12 heterozygous deletion
- 14. Ex) VarSeq NGS data output: LDLR Exon 7 duplication
- 15. Ex) VarSeq NGS data output: LDLR Exons 2-6 duplication
- 16. Results Sensitivity: 100% Specificity: 100% MLPA Result NGS + VarSeq Result Concordance True Positives 38 False Positives 0 False Negatives 0 True Negatives 350 Positive Negative DiploidCNV Iacocca MA, Wang J, et al. J Lipid Res. 2017
- 17. Implications • Use of a single platform (NGS) for detection of both small and large-scale DNA variants • Reduced costs, resources, analysis time associated with the routine molecular diagnosis of FH - MLPA: $80 USD per sample - $31,000 USD for this cohort of 388 samples • Expanding CNV screening to all FH-associated genes on a given NGS panel at no extra cost LipidSeq: APOB, PCSK9 and LDLRAP1, APOE, STAP1, ABCG5/8, LIPA further accounting for all genetic abnormalities capable of defining FH cases
- 18. Future Directions • Novel CNV screening in additional FH-associated genes
- 19. Conclusion • FH is the most prevalent monogenic disorder worldwide affecting ~1 in 250 individuals • DNA testing increasingly becoming a central part of diagnosis; current procedure often includes targeted NGS followed by MLPA • In analysis of 388 FH patient samples, there was 100% concordance in LDLR CNV detection between MLPA and NGS method • Suggests MLPA is dispensable, significantly reducing associated costs, resources, analysis time • All genes on a given NGS panel assessed for CNVs concurrently; allows for novel CNV screening in additional FH genes at no extra cost - promoting more widespread assessment of CNVs across diagnostic laboratories - potential for discovery of novel genetic mechanisms for FH - increasing molecular diagnostic yield
- 20. Acknowledgements Dr. Robert Hegele Dr. Jian Wang Dr. Henian Cao Jacqueline Dron Adam McIntyre John Robinson David Carter Jenn Biltcliffe Rosettia Ho Allison Dilliott Ericka Simon Brooke Kennedy Matthew Ban Blackburn Cardiovascular Research Laboratory
Editor's Notes
- #4: What is FH and why is it important
- #5: ~90 small scale variants ~10 large scale copy number variants; deletions and duplication spanning whole-exons In recent times FH has really moved toward the … The most relevant gene to assess in this effort is the LDL receptor gene, where 90% of FH-causing mutations occur. (CLICK) Over the years, DNA sequencing studies have taught us that about 10% of patients with LDLR mutations carry these large-scale copy number variants, or CNVs -and in this context CNVs are defined as whole-exon deletions or duplications. Currently, the routine procedure for diagnostic labs is the use of 1) NGS panels for the detection of small –scale DNA variants..PLUS MLPA for the detection of these large-scale copy number variants…and this is because until now NGS techniques have been insensitive to CNV detection
- #7: So my objective was to determine if we could apply bioinformatics tools to existing NGS data to detect these CNVs, thus removing the need for secondary
- #8: -so for this study we included 388 individuals with clinically suspected FH -And as mentioned earlier, the routine procedure for molecularly diagnosing FH includes NGS followed by MLPA, which is what we also do in our lab -and then retrospectively, we can go back into this existing NGS data, applying bioinformatics to see if we can detect the same CNVs detected by MLPA
- #9: We use the illumine ..
- #10: 4.3.7 Filter the variants that have been called based on their overlap with the targeted panel’s target regions as specified by the Browser Extensible Data (BED) file, allowing only variants occurring within the genomic regions selected for the targeted NGS panel to be retained. Note: The BED file will be unique to the targeted NGS panel that is being utilized, based on the regions of the genome that the panel is able to cover.
- #12: OK so with MLPA we detected CNVs in 38 of 388 patients, about 10% -the most common CNV was a heterozygous deletion of the promoter and exon 1 which was found 22 times -but there were CNVs detected that affected pretty much every one of LDLR’s 18 exons (CLICK) -ok, so then with our bioinformatics approach we were able to detect all 38 of these CNV events as well
- #17: And furthermore, the 350 samples that were negative for CNVs with MLPA, were also all negative for CNVs with our bioinformatics approach Taken together, in a 2x2 contingency analysis when considering MLPA as the gold standard or reference standard …these results translate to a TEST sensitivty AND specifity of 100% each
- #18: - ~100 CAD per MLPA sample, $38,000 for this cohort of 388 individuals - No MLPA available for these genes
- #19: Despite being rare, they have long remained uninvestigated up until this point , cost-benefit assumed low